Unverified50% confidenceBenchmarkExact time
In the experiment, the two size points of 2.8KB and 4.1KB were executed 8 times in total, with empty-output rates of 0% each, average output tokens of 9,096 and 7,913 respectively, and convention hit rates of 4/4 each.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTaking a 7B parameter model as an example, each additional 1K tokens of context consumes approximately 0.5–1 GB of GPU memory.74% similarUnverifiedThe naturally bloated 9.2KB skill file had an empty-output rate of 50%, averaged 14,766 output tokens, and had a convention hit rate of 2/4; compared with the manually filled 9.5KB file (25% empty-output rate), the naturally bloated file had a higher failure rate.74% similarUnverifiedOn the same machine and model, the decoding speed difference between low power mode and high power mode is 1.4 to 3.8 times; for example, running Qwen 3.5 35B on Ollama gives 51.9 tokens/s in low power and 107.1 tokens/s in high power.71% similarUnverified以隐藏维度 8192 的 70B 模型、batch size = 1 为例，每生成一个 token 穿越 PCIe 的总数据量只有几 MB。71% similarUnverifiedWhen running 4-bit DeepSeek R1 on the M3 Ultra, a 1,000-token context generates approximately 19 tokens per second using 392GB, dropping to approximately 15 tokens per second using 410GB at a 16,000-token context.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1062633API
curl https://kongchang.com/api/v1/knowledge/claims/1062633MCP
get_claim(id=1062633)